Is Output shape available in MyNewContribOp::Compute()?

[Dan5002]

I see that Input shapes are available in the context:
Status MyNewContribOp::Compute(OpKernelContext* context) const {
auto X = context->Input(0);
auto& dims = X->Shape();

But I would like that to be true for Outputs as well.

I don't want to hard code it:
TensorShape output_shape_nms({1, 15130, 4});
Tensor* output_to_nms = context->Output(1, output_shape_nms);

The shape is available when I graph my model, as my outputs are inputs to other nodes. See screenshot of my model displayed with Netron.

Is there a way to parse the graph and find the expected shape of my output? I don't see Output shape in the context.

[Dan5002]

See comment in bold below. This is just a restatement of the question in the prior post:

Is there a way to get the 'given tensor shape' from the graph, within the Compute() method (using context, I suppose)?

From include/onnxruntime/core/framework/op_kernel.h:

// Fetch output (non-tensor) with specified index.
template
T* Output(int index) {
if (index < 0 || index >= OutputCount())
return nullptr;

OrtValue* p_ml_value = GetOrCreateOutputMLValue(index);
return p_ml_value ? p_ml_value->GetMutable<T>() : nullptr;

}

// In the case that memory allocation has not been done for an output tensor,
// The memory allocation will be done on-the-fly with given tensor shape.
// Return nullptr if the output is an unused optional output.
Tensor* Output(int index, const TensorShape& shape);

[pranavsharma]

Currently there's no way to fetch the output shape from the graph using OpKernelContext.